How moving to the cloud shifts your liability (explained with cars)
One of the unappreciated benefits when moving to the cloud is that you're paying for someone else to take the wheel. We explain with lots of car metaphors.
Jul 11, 2023 • 6 Minute Read
Cloud computing in the workplace is as common as your corner coffee shop, or the smart phone in your pocket — in fact, these are likely connected to the cloud as well! But when it comes to businesses, not a lot of leaders are turning cloud use into customer value. In Pluralsight’s State of Cloud 2023, only a quarter of leaders said their cloud initiatives are driving value. Yikes!
That’s a big deal, particularly if it’s coming with a big price tag. Part of this is due to poor implementation processes, like lifting and shifting your existing infrastructure into the cloud and expecting it to magically manifest value. Given that the same report found 69% of leaders didn’t have a clearly defined cloud strategy, I’d say that fits.
However, one thing that I find is frequently overlooked when it comes to cloud adoption is the benefit of shifting your liability—away from you and to someone else, I mean. By overlooking this benefit, people don’t make the most of what they can do once they’re in the cloud that can make their lives easier.
In this article, I’m going to explain the different levels of delegating responsibility to your cloud provider, and how this benefits both your organization and anyone working on managing your infrastructure.
The levels of cloud liability
When it comes to delegating responsibility to your cloud provider, I find it’s helpful to use the analogy of self-driving cars. Self-driving cars come in different levels depending on how much autonomy you give them, with level zero being no automation, and level five being completely automated (The steering wheel is optional!)
You can think of shifting responsibility to your cloud provider in exactly the same way. So, let’s jump in!
Level zero: No automation (Manual VMs)
Your organization has some VMs in the cloud, but everything’s manual. Need more VMs? You spin them up. Need to save money? You shut some down. Your hands are on the wheel and your feet are always on the pedals, just like an old Volkswagen Beetle.
Level one: Shared control (Observability/Monitoring, VM Scripting)
At this level, you’ve got some nice monitoring tools (e.g. AWS Cloudwatch, Azure Monitor, Google Cloud Monitoring) and have access to powerful scripting via the CLI or SDK. These are nice quality of life improvements, but you’re still making things happen, just more easily than before. In driving terms, it’s like you’ve got a fancy HUD and access to cruise control.
Level two: Hands off (VM Autoscaling)
At level two, the cloud providers are doing stuff for you. You no longer have to add VM instances or shut them down depending on the load or time of day. But don’t be fooled, the responsibility is still on you! If you have misconfigured the autoscaling, then your service will suffer (Either in poor performance because you’ve scaled too low, or expenses because you’ve scaled too high).
This is still “hands off” because you’re not touching it, but everything that happens is still 100% on your radar, and you need to be okay with it. In terms of cars, this would be the Tesla’s FSD Beta: it does everything for you — navigation, steering, handling speed, and signaling — but you need to be watching the road and ready to take over. If you don’t, you’re completely liable for what happens.
Level three: Eyes off (Serverless functions)
At first glance, this seems exactly the same as level two. However, the key difference is liability, specifically who pays the cost for any problems with scaling.
At this level, in the cloud, your only involvement is to make your functions faster to scale (whether by tweaking your code, pinning instances, etc). But all the extra work the cloud providers do behind the scenes to pre-warm containers for your functions, and all that good stuff? That’s on them. If they scale too high, they pay the extra cost, and if they scale too low, you complain and threaten to move to another cloud. In the cloud, services like this would be AWS Lambda, Azure Functions, or Google Cloud Functions.
In the world of automated cars, these are cars where you can text or watch a film, and the vehicle will handle any emergency situations like braking. There are still some rare situations where you might be required, so you can’t leave the actual driver’s seat.
Level four: Mind off (Serverless services and integration)
The services scale internally and transparently to you. The only way you know that it has scaled is in the bill. You still have to hook everything up and you pay for usage, but there’s not even any way for you to control the scaling. In the rarest of circumstances, there may be a leak in the abstraction that makes you remember that scaling happens under the hood, but that is considered a deficiency in the service and they work to fix it. Examples: S3/GCS/ABS, SQS, SNS, BigQuery.
In short, you can safely go to sleep, or even leave the driver’s seat, so long as you stay within the limits of your geofenced area.
Level five: Steering wheel optional (SaaS)
There’s nothing to do except use the software. Scaling is completely behind the scenes, and the bill is predictable. An example of this is Gmail! If you can replace your custom cloud software system with something as robust as Gmail, go for it. But for most custom situations — much like a car without a steering wheel that can drive anywhere — this is currently unattainable.
Lowered liability is a significantly underrated benefit of cloud computing
Jumping from level two to three is a dramatic shift, moving the liability from you to the provider. If something bad happens, you can blame them, charge them, sue them, or whatever the appropriate response is. Sure, you might be impacted by what happens — like you might not make it to your meeting on time if your self-driving car gives up and pulls over — but you should not have to pay the cost of any very unlikely accident it may cause.
Shifting responsibility allows for taking advantage of scale
Once you’ve shifted things up a few levels, your responsibility lessens, and you can spend time taking advantage of the scale that cloud providers can leverage. They have things like algorithmic prediction and distribution for managing resources that you don’t have to develop, manage, or pay for—you can just simply use them. This is just one more reason that the myth that “the cloud is just someone else’s computer” really misses the true benefits of cloud computing.
When you think of the time that it takes you to do things at level zero or one, the extra cost is quite substantial, and it massively reduces your agility. Agility is frequently called out as the biggest benefit of cloud computing, so missing out on it means missing out on huge business value.
